A new disaster victim identification management strategy targeting “near identification-threshold” cases: Experiences from the Boxing Day tsunami

Forensic Science International 250 (2015) 91–97

Contents lists available at ScienceDirect

Forensic Science International journal homepage: www.elsevier.com/locate/forsciint

A new disaster victim identification management strategy targeting ‘‘near identification-threshold’’ cases: Experiences from the Boxing Day tsunami Kirsty Wright a,*, Amy Mundorff b, Janet Chaseling c, Alexander Forrest a,d, Christopher Maguire e, Denis I. Crane a,f a

School of Natural Sciences, Griffith University, 170 Kessels Road, Nathan, Queensland 4111, Australia Department of Anthropology, The University of Tennessee, 250 South Stadium Hall, Knoxville, TN 37996, USA c School of Environment, Griffith University, 170 Kessels Road, Nathan, Queensland 4111, Australia d Queensland Health Pathology and Scientific Services, 39 Kessels Road, Coopers Plains, Queensland 4108, Australia e District of Columbia Department of Forensic Science, Washington, DC 20024, USA f The Eskitis Institute for Drug Discovery, Griffith University, 46 Don Young Road, Nathan, Queensland 4111, Australia b

A R T I C L E I N F O

A B S T R A C T

Article history: Received 10 December 2014 Received in revised form 23 February 2015 Accepted 10 March 2015 Available online 18 March 2015

The international disaster victim identification (DVI) response to the Boxing Day tsunami, led by the Royal Thai Police in Phuket, Thailand, was one of the largest and most complex in DVI history. Referred to as the Thai Tsunami Victim Identification operation, the group comprised a multi-national, multiagency, and multi-disciplinary team. The traditional DVI approach proved successful in identifying a large number of victims quickly. However, the team struggled to identify certain victims due to incomplete or poor quality ante-mortem and post-mortem data. In response to these challenges, a new ‘near-threshold’ DVI management strategy was implemented to target presumptive identifications and improve operational efficiency. The strategy was implemented by the DNA Team, therefore DNA kinship matches that just failed to reach the reporting threshold of 99.9% were prioritized, however the same approach could be taken by targeting, for example, cases with partial fingerprint matches. The presumptive DNA identifications were progressively filtered through the Investigation, Dental and Fingerprint Teams to add additional information necessary to either strengthen or conclusively exclude the identification. Over a five-month period 111 victims from ten countries were identified using this targeted approach. The new identifications comprised 87 adults, 24 children and included 97 Thai locals. New data from the Fingerprint Team established nearly 60% of the total near-threshold identifications and the combined DNA/Physical method was responsible for over 30%. Implementing the new strategy, targeting near-threshold cases, had positive management implications. The process initiated additional ante-mortem information collections, and established a much-needed, distinct ‘‘end-point’’ for unresolved cases. ß 2015 Elsevier Ireland Ltd. All rights reserved.

Keywords: Boxing Day tsunami Disaster victim identification (DVI) Forensic biology Dental age estimation Fingerprinting Thai Tsunami Victim Information (TTVI) operation

1. Introduction On 26 December 2004, an earthquake measuring 9.1 on the Richter Scale triggered a massive tsunami responsible for more than 280,000 deaths in thirteen countries. In Thailand 5395 victims were recovered, including approximately 2400 foreign tourists representing 41 nationalities [1–3]. An identification operation commenced immediately, resulting in local authorities releasing

* Corresponding author. Tel.: +61 07 37357773. E-mail address: k.wright@griffith.edu.au (K. Wright). http://dx.doi.org/10.1016/j.forsciint.2015.03.007 0379-0738/ß 2015 Elsevier Ireland Ltd. All rights reserved.

approximately 560 bodies based on visual identification by their families. Thai forensic teams released an additional 1151 bodies, 111 of which were based on dental examination and the rest based on physical and property examination [4]. The Thai identification effort was later joined by international experts from over 30 countries working as part of the Thai Tsunami Victim Identification (TTVI) operation in Phuket, Thailand. By the commencement of the international DVI operation in January 2005, there were 3679 unidentified bodies, approximately half of which were thought to be Thai locals [1]. The TTVI operation followed the DVI management strategy outlined by INTERPOL [5]. Plass Data’sTM ‘DVI System International’ was used to record, store and electronically

92

K. Wright et al. / Forensic Science International 250 (2015) 91–97

search ante-mortem data (AM) and post-mortem data (PM) contained in INTERPOL DVI forms and an Automated Fingerprint Identification System (AFIS) was used to search fingerprints [6–9]. This article examines the implementation, capabilities, and contributions of a novel identification approach executed during a complex DVI operation. A new DVI management strategy, specifically targeting cases yielding below the established DNA identification threshold, was implemented in response to the significant decline in single-modality identifications. The approach was designed to achieve positive identifications for complex cases and to identify cases lacking the necessary information to establish an identification, to request additional data collection. 1.1. Background Typically, primary (DNA, Dental, and Fingerprints) and secondary (property and physical) identification teams work independently of each other and of police investigators. The teams search for matches between AM and PM records that reach nominated reporting criteria based on international standards and that are set by the command structure within each DVI operation [5]. The separation of the identification teams in DVI mimics the approach used by many forensic experts for criminal cases, which is designed to eliminate cognitive and confirmation bias and promote independent interpretation of forensic analysis. This was the approach used at the TTVI, with a Data Mining Team searching data for matching secondary identifiers (property and physical identifiers such as scars and tattoos). Following the tsunami, AM records were received from 42 countries and the PM records were generated by a rotating staff from at least 30 countries as part of the TTVI mortuary operations. As seen with other DVI operations, identifications can be quickly established when accurate and complete AM and PM data are available [10,11]. Therefore, the traditional DVI strategy worked well, particularly during the early part of the operation. This is clearly evidenced during the months of February and March 2005, when the majority of identifications were established using a single primary method, (93.02% and 87.26%, respectively) (Fig. 1). By April 2005, 90.18% (1057 of 1171) of the dental identifications as a stand-alone method had already been achieved. This was followed by a rapid and precipitous decline in the proportion of cases being resolved by a single primary identification method. Highlighting this drop further, overall single-modality identifications declined in the month of December 2005 to only 20.33% (12 of 59) (p < 0.0001). By 19 January 2006, a combination of identification methods was necessary to establish nearly 30% of all identifications [1]. On 9 January 2008, a total of 3761 AM records and 3696 PM records had been entered into their respective databases. However,

Fig. 1. Percentage of total identifications each month achieved with stand-alone methods.

the forensic data available in the AM records was limited, with only 58.1% containing DNA evidence, 52.6% Dental evidence and 41.3% Fingerprint evidence [12]. The low percentage of AM records with forensic evidence helps explain why stand-alone methods of identification could not be used to resolve a number of cases. Missing or incomplete AM data can be attributed to a number of factors, many of which have been experienced during other DVI operations [13,14]. The reasons for incomplete AM data confounding identification efforts specific to this disaster include:  Lack of AM dental records for Thai locals. Petju reported that only 18.1% of missing Thais had dental charts and only 0.8% had dental X-rays [15]. This compares to missing Europeans, of whom 94.4% had dental charts and 75.5% had dental X-rays. Of the 18.1% of Thai victims with dental records, only 7% were used to establish identity [12].  Loss of reference samples due to the tsunami’s destruction, including local AM dental records and identifying personal effects for the numerous victims on the beach at the time of the event [16].  Lack of AM dental records for victims originating from Myanmar. This may be attributed to limited access to dental care or because family members were hesitant to come forward and supply AM information if their missing relative was in the country working illegally.  Lack of DNA reference samples from family members of victims, many of whom were also victims. In some cases entire families were missing.  Lack of fingerprint reference samples for local Thai children too young to have their fingerprints recorded on national identity cards [1].  Difficulties in obtaining reliable AM fingerprints from missing foreign children [1].  A ‘preferential approach’ to the collection of AM material adopted by many countries early in the operation. For example, some countries targeted data for dental identification; while others decided DNA would be the best method and chose not to supply fingerprint and dental data [17]. Using the standard DVI approach, investigators struggled to increase identifications beyond a certain point. Instances of incomplete data, or data compromised by data entry errors, prevented matches. Additionally, segregating the identification teams, a typical DVI strategy, led to duplicated efforts between staff rotations. Each national DVI team typically had a staff rotation between two and five weeks, where new experts would join a team, replacing those returning home. The newly rotating individuals would begin to search for possible matches among cases previously discounted within their group or discarded by another primary team because there was no mechanism in place to record confirmed exclusions. Complex cases often did not continue to progress when handed off to a new rotation and critical information was not always transferred between teams when an investigator finished a rotation. While a ‘Targeted Request’ form was in place, the process was not efficient or systematic and was not structured to leverage and integrate all possible information from the primary and secondary identification teams. Moreover, the DVI strategy was based on searching for AM and PM matches; exclusions were not consistently or systematically recorded. There was no established ‘‘end-point’’ for cases lacking sufficient AM data. This was a critical void in the identification process given the large number of AM cases with incomplete records (DNA 41.9%, Dental 47.4% and Fingerprints 58.7%) [12]. Although a paper-based process existed to allow an investigator to request additional AM information, this was typically limited to a single discipline on a case-by-case basis and not adequately

K. Wright et al. / Forensic Science International 250 (2015) 91–97

implemented or rigorously followed-up. In effect, these cases remained in limbo and were repeatedly investigated without resolution due to their lack of identification potential. Each team recognized cases that they categorized as a ‘possible’ or ‘probable’ match but the evidence was not strong enough to progress to the Reconciliation Board. At such a point in time, there was no formal process to direct these cases for systematic and independent comparison across other Identification Teams, to bolster existing evidence and build a stronger basis for identification. Each Identification Team had a small piece of the ‘identification jig-saw puzzle’ that, when combined, may have established an identification, but operational efficiencies were not in place to allow this to happen. As a result, more complex cases lingered, still unidentified, and single-modality identifications declined. 2. Materials and methods A more efficient system was necessary to strategically target information from cases that fell just below the reporting threshold for identification. An approach was needed that could systematically select those cases for targeted investigation. In June 2005 a new approach was tested with promising results. Twelve DNA kinship matches that were below the 99.9% reporting threshold were selected for additional investigation. Five of the targeted cases resulted in new identifications through additional fingerprint data. The following month, a new, targeted identification strategy focusing on all ‘‘near-threshold’’ DNA matches was formally implemented into the TTVI (see Fig. 2). The new approach focused on ‘‘near-threshold’’ DNA matches. However, this strategy could be adapted so other identification methodologies, such as dental or fingerprint analysis, could also be

93

used to prioritize high probability matches. Cases selected for investigation were progressively filtered through analyses to exclude a match, provide strength to a weak match or to target the case for additional AM information collection. The individual elements of this strategy were as follows: 1. Defining the List to be Investigated: The DNA Team generated a daily list of ‘‘possible to probable’’ DNA kinship matches with a posterior probability of 90.0–99.8% (the positive identification threshold was set at 99.9%). The list was electronically uploaded to a database specifically designed for these special investigations. 2. Property and Physical Information Evaluation: The Investigations Team selected a case from the database and evaluated its associated property and physical information (e.g., height and personal property) in order to quickly exclude cases and narrow down the list of potential matches. The Property and Physical Information Evaluation was strategically positioned as the first investigative-filter because this rapid exclusionary process reduced the workload for the succeeding team. Dental and fingerprint investigations required specific experts who, when compared to law enforcement personnel, were a limited resource at the TTVI. When exclusion was not possible, the evaluation was recorded in the database and the case was progressed to the Dental Team. a. Concurrently, newly discovered individualizing information (e.g., inscribed wedding ring, scars, and tattoos) was transmitted to the DNA Team to be used to further adjust the prior probabilities for that specific case. For example a uniquely inscribed wedding ring recovered from a victim could be used to reduce the prior probability from 1/3000 to 1/2.

Fig. 2. The near-threshold strategy used at the TTVI.

94

K. Wright et al. / Forensic Science International 250 (2015) 91–97

b. Potential matches were recalculated using the updated prior probabilities, which often led to the posterior probability reaching the 99.9% threshold. When this occurred the identification method was designated as combined ‘DNA/ Physical’. 3. Dental evaluation: The next evaluation was an in-depth dental re-examination. Matches were forwarded to the Reconciliation Team and exclusions were returned to the Investigations Team. When exclusion was not possible, the case was progressed to the Fingerprint Team. a. Again, additionally acquired data, such as dental age assessment, was transmitted to the DNA Team for adjusting prior probabilities between the adult and child sub-categories. b. Potential matches were reassessed using the updated prior probabilities. 4. Fingerprint Evaluation: The Fingerprint Team manually compared hardcopies of AM and PM fingerprints (rather than matching with AFIS). Fingerprint analysis was intentionally placed at the end of the process to limit the number of cases necessitating a labor-intensive manual comparison. Matches were progressed to the Reconciliation Team and exclusions were returned to the Investigations Team. The Fingerprint Team also flagged cases that could not be evaluated due to insufficient AM or PM prints. a. Cases that could not be definitively excluded, or could not be identified due to lack of AM data or poor quality data, were targeted for additional AM information collection. 5. Reconciliation: The Reconciliation Team evaluated all possible matches and assessed whether the evidence was sufficient to scientifically support identity. The Reconciliation Board took the official decision for identification. Throughout the process, the Investigations Team managed the database, continually updating the system to reflect each subsequent evaluation. Details specific to each case were recorded in the database to track a case’s progress, record exclusions, confirm identifications, and identify cases requiring additional data. The database supplemented ‘DVI System International’, which lacked these capabilities. The database also provided effective and efficient case management, and a means to maintain independence between teams. Utilizing data from multiple databases, cases processed through the near-threshold strategy were evaluated as to whether or not the process contributed to identifications. The specific databases utilized during this analysis included:

potentially underestimating the reported number of cases assisted by the near-threshold strategy). This was an Excel spreadsheet created and used internally by the Reconciliation Team to record the AM and PM number of each confirmed identification, the date the case was accepted by the Identification Board, which identification method/s were used, the age, sex and nationality of the victim. Evaluation of the ‘DNA Investigations’ database revealed data entry inconsistencies among operators. Progress summaries varied in detail and instances where data field entries were ambiguous or contained no data (e.g., result or date of investigation) were noted. Therefore, sole reliance on this database was not sufficient for evaluation necessary to attribute a successful identification to the near-threshold strategy. The specific criteria outlined below, along with additional information from the other two databases, allowed for the evaluation of ambiguous cases for inclusion or exclusion into the data set. To qualify as an identification that was assisted by the nearthreshold strategy, the case must have been initially flagged by the DNA Team as a possible kinship match, unidentified at the time of investigation and met at least one of the following criteria: 1. Comments on the ‘DNA Investigations’ database indicated that one of the identification teams investigated and confirmed the possible match (either as stand-alone or part of a combined identification). 2. Comments on the ‘DNA Investigations’ database indicated that the case could not be excluded and was progressed for further examination by the Dental or Fingerprint Teams, and then appeared on the ‘Reconciliation’ database within two weeks of being investigated. 3. Comments on the ‘DNA Investigations’ database indicated that the AM case required further information to progress and the case was subsequently identified by this additional information. 4. Comments on the ‘DNA Investigations’ database indicated physical, property or dental information was used to support the DNA match, or that comments returned to the DNA Team allowed for adjusted prior probabilities leading to an identification in the ‘Reconciliation’ database. Cases were not considered for inclusion in the dataset if there was insufficient information or if the DNA match reached the 99.9% posterior probability threshold before investigation. 3. Results

1. The ‘DNA Investigations’ database, created July 2005 at the TTVI, which contained new case entries through to 4 October 2005. The ‘DNA Investigations’ database was an Excel spreadsheet developed and maintained by the Investigations Team. It contained the AM and PM case number for the presumptive DNA kinship match, information on targeted searches that included evaluations of height, age, personal property, and dental and fingerprint comparisons. The database was used to record and track exclusions and identifications, and cases requiring additional AM data. 2. The TTVI ‘AM and PM DNA Matches’ database created April 2005 with entries until 23 September 2005. This was an Excel spreadsheet created and used internally by the DNA Team to record the AM and PM case number of all DNA matches, the posterior probability for the match and whether the cases progressed to the Reconciliation team via a DNA report. 3. The TTVI ‘Reconciliation’ database created in January 2005 with entries through to 12 December 2005 (access to the TTVI reconciliation data after 12 December 2005 was restricted,

A total of 792 cases were examined as part of the new strategy. All cases were initiated for investigation because the kinship DNA matches fell just below the reporting threshold. The nearthreshold strategy revealed immediate success, with 49 new identifications achieved during the first full month of implementation (Table 1). A total of 111 new identifications were established which amounts to 10.31% of all TTVI identifications during the strategy’s 5-month implementation period (23 July 2005 and 12 December 2005). Identifications were comprised of 87 adults and 24 children (16 years and under); 97 of the 111 new identifications were Thai nationals; three were Swedish nationals, two each from China, Finland and Myanmar and one each from France, Germany, Japan, Russia and the United States. Even though the cases targeted for investigation under this strategy were selected because they were all just under the threshold for DNA identification, this did not preclude other modalities from establishing identifications or limit identifications to DNA combined with another method. The strategic flow

K. Wright et al. / Forensic Science International 250 (2015) 91–97 Table 1 Number of identifications assisted with the new strategy. All IDsa

Month (2005) July August September October November December Total a

New strategy IDs

% of new strategy IDs

227 185 167 228 211 59

1 49 25 26 8 2

0.44% 26.49% 14.97% 11.40% 3.79% 3.39%

1077

111

10.31%

Identifications.

developed through this process allowed other identification teams to establish new, stand-alone identifications with information not previously noted due to data quality issues. For example, the Fingerprint Team recorded 66 stand-alone identifications and contributed to another ten combined identifications (Table 2). The process also allowed other identification teams to provide supportive data to bolster the original DNA match. A combined DNA/Physical identification (‘physical’ included property and medical/autopsy information) was achieved on 35 occasions, and the remaining 10 identifications were assisted through different combinations of methods. Twenty-one new identifications resulted from additional AM data collections. Part of this strategy focused on identifying cases with insufficient AM material necessary to resolve identification. Once recognized, these cases were brought to the attention of Liaison Officers who worked with police in the victim’s country of origin. Where possible, requests for additional information were accompanied by specific advice on the type of AM information to collect. For example, additional family DNA reference samples may be requested to boost the kinship probability to the reporting threshold. 4. Discussion 4.1. Resolution of difficult cases using the near-threshold strategy Thai nationals proved difficult to identify owing to the lack of AM dental records. More reliance was placed on fingerprint and DNA analysis, however, access to sufficient AM DNA reference samples was also problematic. The near-threshold strategy proved successful for cases lacking adequate AM data. The significant number of Thai victims identified under this program reflects the benefits of an alternative targeted approach. Evaluating physical and property evidence contributed strong secondary information leading to nearly one third of the new identifications. The objective of having this analysis at the beginning of the process was to resolve near-threshold DNA cases with secondary information from physical characteristics or property to minimize the use of dental and fingerprint resources. To increase operational efficiencies, individuals with varying Table 2 Number of identifications assisted, by method, using the near-threshold strategy. Method

IDsa

Fingerprints DNA/Physical Fingerprints/DNA Fingerprints/DNA/Physical Fingerprints/Physical Fingerprints/Dental

66 35 3 3 3 1

a

Identifications.

95

backgrounds could be trained to evaluate secondary identifiers (property and physical characteristics) when access to dental or fingerprint specialists was limited. Dental comparison is only successful in rapidly identifying large numbers of victims when good quality AM and PM records are available. Individuals without sufficient AM dental records are not easily identified by this approach. For example, the Dental Team identified only 2.0% of Thais (given only 18.1% of missing Thais had dental charts and only 0.8% had dental X-rays), but identified 76.4% of Europeans and 76.5% of North Americans (who had 94.4% and 88.2% of dental charts and 75.5% and 76.5% of dental X-rays, respectively) [15]. However, the new targeted approach facilitated dental evaluation even when the AM data was limited. Key informative details, such as dental age estimates, were gleaned from the Dental Team’s analyses. The PM cases involving children provided the most value in terms of accurately estimating the age of the victim due to specific dental development patterns [1]. In addition, instances of inconsistent age estimations recorded on the autopsy form, individuals too broadly nominated as either adult or child, and cases with incorrect age estimations were revealed through this evaluation. James [2] has discussed specific quality issues of dental records, including ambiguous AM record collection by general dentists and occasionally police officers rather than experienced forensic dentists, a lack of AM radiographs from public clinics, poor quality AM radiographs, and poor dental examination record taking and record keeping. The Dental Team found that the reproduction of original AM radiographs by photocopying, photographing, faxing or scanning, removed detail necessary for matching. Keiser reported that of 106 TTVI AM dental records examined for quality control, only 49% were accepted, with the remaining 51% returned because the data were either incomplete or of an unacceptable standard [18]. The Tsunami Evaluation Working Group reported that such reproductions were frequently flawed and stated ‘‘each step away from the original document or radiograph introduces the possibility and the probability of error, resulting in a significant waste of time and also the probability of non-identification’’ [1, p. 110]. The PM dental records were of a higher standard than AM dental records. Keiser examined 78 PM records for quality control and determined that 92% of dental charting was considered good quality, 85% of radiographs were considered good quality and 96% of photographs were considered good quality [18]. The near-threshold strategy revealed a large number of potential fingerprint matches that were not detected by the automated AFIS system. In these instances, data quality issues prevented electronic matches. For example, poor quality AM or PM prints, inconsistencies in the relative sizes of the scanned prints, or the way in which the data were electronically processed for transfer to the TTVI from the host countries. Fingerprints were scanned at a resolution of 500 dpi before being entered into AFIS, but such resolution did not allow for electronic matching of certain finer detail, particularly as found in fingerprints from children. At that point, the software did not allow for an increase in resolution. Instead, fingerprint examiners relied on hard copy prints and a magnifying glass to manually detect the finer details needed for some matches [1]. This was not a practical approach for comparing thousands of records; therefore the Fingerprint Team relied on targeted requests from the near-threshold strategy to progress more complex cases affected by data quality issues. In addition, this strategy revealed a few instances where the AM fingerprints had not been uploaded to AFIS, so those cases were targeted for collection, upload and comparison. The quality of AM fingerprints also varied, as some countries collecting prints in the homeland of missing foreign tourists were not aware of the quality standards needed. Another issue was the electronic transmission of AM fingerprints without an appropriate

96

K. Wright et al. / Forensic Science International 250 (2015) 91–97

scale, rendering the prints incompatible for matching against PM fingerprints in AFIS [17]. The impact of the AM data quality issues was recognized in the INTERPOL Tsunami Evaluation Working Group report, which highlighted that speedy and successful identification can only be achieved with quality AM data relating to the missing. The group recommended ‘that INTERPOL oversee the development of standard operating procedures and minimum data standards for the submission of quality ante mortem data by foreign countries’ [1, p. 96]. Data quality issues were not unique to the Boxing Day tsunami DVI operation; other large-scale disasters have confronted similar problems that impacted successful identifications. Hennessy examined six areas of possible quality failure in the collection of AM DNA records for the World Trade Center Attacks [19]. The six areas of possible quality failure were inaccurate data from the informants, misunderstandings between informants and interviewer, uneven DNA training for interviewers, information incorrectly recorded by interviewers, lack of data entry standardization and handwritten interviews. He speculates that if 99% accuracy is achieved at each point of possible quality failure, then the final record will only be 94% accurate; this strongly advocates for a stringent administrative review process. Donkervoort examined the data accuracy for the collection of AM family reference DNA from Hurricane Katrina and reported on the three fields considered to be critical information for achieving DNA identification, namely: number of missing relatives, relationship of the donor to the missing relative and verification of relationship through pedigree. She reported that only 30% of the forms evaluated for the study were completely accurate [20]. 4.2. Improvements in operational efficiencies The near-threshold strategy introduced operational efficiencies during the mid to later stages of the TTVI operation. It is necessary to implement such management strategies, particularly when evaluating difficult cases, to minimize redundant efforts and clearly define and endpoint. Implementation of the new nearthreshold strategy improved operational efficiencies by:

which allows for these types of liaisons to exist, even within a large-scale disaster. Software systems used to rapidly compare large quantities of data are critical for DVIs involving high numbers of deceased. However, poor quality data may lead to a false exclusion, or worse, an incorrect identification. Dailey highlights the importance of quality control for AM and PM dental records in a DVI. He also warned against the involvement of over-confidence in inexperienced operators and negative effects of staff ‘burnout’ in these circumstances [21]. At the TTVI, AM and PM data quality issues resulted from differences in operating procedures, non-standardized terminology, inexperienced personnel, fatigue, the use of disparate databases, issues arising from the reproduction of original dental and fingerprint records, problems during electronic transfer, and manual entry errors [1,22]. Along with incomplete data, the decline in single modality identifications was also attributed to poor data quality. A number of measures were implemented to increase data quality. These included a Quality Assurance Team at the mortuary sites; an AM and PM Data Quality Team to address data entry; and the introduction of standard operating procedures into the work practice of individual teams. A ‘Final Inventory Protocol’ was also undertaken at the mortuaries to re-examine unidentified victims in an attempt to detect errors, collect additional data, and validate existing information [1,2,9,23]. Ultimately, implementing the near-threshold strategy resulted in swift resolution of difficult cases. Creating a process with a discrete end-point led to the targeted acquisition of additional information. In turn, this facilitated swift identifications ultimately conserving time and monetary resources. The value of a systematic end-point was recognized following the World Trade Center Human Identification Project by Biesecker et al. [24]. One of their seven recommendations for future disaster management planners states, ‘‘Criteria for determining end-points should be designated early in the identification process’’ [24, p. 1122]. Therefore, the duration of large-scale DVI operations could be considerably reduced through implementing this type of approach.

5. Conclusions 1. Ending redundant evaluations of the same cases within and between teams; 2. Excluding ‘possible’ and ‘probable’ DNA matches, thereby saving DNA resources; 3. Documenting the outcome of each case investigation; 4. Flagging cases that could not proceed to an identification without additional AM information, thereby establishing an end-point to the repeated evaluations; 5. Targeting the collection of additional specific AM data for individual cases; 6. Targeting ‘possible’ DNA matches that otherwise would not have been reported; 7. Systematically addressing data quality issues through identification of inaccurate records and collection of better quality information.

Constantly re-examining cases without reaching a resolution costs valuable time and resources, and delays the inevitable request for additional AM data necessary to achieve identification. Smaller DVI operations allow for easy liaison between identification teams and the Reconciliation Team. These targeted discussions can then lead to a request for additional AM material. This simply was not possible at the TTVI because of the large number of victims (3679 by January 2005) and frequently rotating personnel. The implementation of the near-threshold strategy is a mechanism

To maximize efficiency and identifications, a near-threshold strategy should be implemented to augment a traditional DVI approach, particularly when stand-alone identifications begin to decline. Management personnel can utilize this type of strategy to generate a target list of potential matches, which can be filtered through identification teams while systematically recording investigative outcomes. The traditional DVI approach allows for rapid identification by stand-alone methods, provided that access to good quality AM and PM data is available. However, the specific characteristics of the Boxing Day tsunami (i.e., people on the beach without personal effects, limited access to dental care, entire families’ killed, and mass destruction of homes) resulted in a significant number of cases with incomplete AM data. Conversely, incidents with focused disaster locations such as the events of 11 September 2001 (the Twin Towers, the Pentagon, a field in Pennsylvania), still allowed police to collect items from the victims’ homes and family members were still available for DNA references. Similarly, following the 2002 Bali Bombings, the police were able to collect items from the victims’ hotel rooms and DNA from living family members. In these disasters, it was the incomplete PM records that caused the identification difficulties due to fragmented and compromised human remains [25,26]. However, the bodies of the tsunami victims were mostly intact and often well enough preserved to obtain fingerprints. Petju reported that of 3750 bodies

K. Wright et al. / Forensic Science International 250 (2015) 91–97

examined, 97.4% (3652) were considered whole bodies and only 2.6% (98) were body parts or bodies without a head [15]. For large DVI operations such as this, where complete and accurate AM or PM data is not always available, a near-threshold strategy should prove useful. Many of the successful identifications achieved through this process were the result of this strategy to identify cases requiring additional AM information for resolution. This approach had a mechanism in place to flag cases unable to progress with the limited information available, resulting in specific requests tailored to each case. Supplementary casespecific information necessary to adjust prior probabilities thus allowing for identification by DNA also resulted from this approach. In addition, implementation of this new strategy helped address data quality problems in both AM and PM information. More specifically, transcription errors not detected during the initial data entry process or problems encountered during data upload into AFIS, both of which hindered identifications. There are many difficulties associated with large-scale DVI operations. More complex cases often necessitate the use of combined methods for identification, to complement stand-alone methods. In fact, there are recommendations that support using combined information for mass fatality identification operations [25]. Prinz et al. [27] recommends that ‘‘Especially if multiple family members are involved, DNA-based identification should whenever possible be anchored by anthropological and/or circumstantial data, a second identification modality, or multiple DNA references’’ [27, pp. 9–10]. A combined identification method can increase identification efficiencies, increase the level of confidence in identification and potentially decrease identification errors. In summary, this article illustrates that a traditional DVI operation benefits from the concurrent implementation of a strategy targeting complex cases. Simultaneous implementation during a large-scale DVI operation can be successful while also maintaining the original organizational structure, rigid scrutiny of matches and specific identification team responsibilities. A minimum of 111 identifications were assisted by the nearthreshold identification strategy during its 5-month implementation period. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.forsciint.2015.03. 007. References [1] INTERPOL Tsunami Evaluation Working Group, http://www.interpol.int/Media/ Files/INTERPOL-Expertise/DVI/INTERPOL-Tsunami-Evaluation-Working-Group (accessed 18.06.14). [2] H. James, Thai Tsunami Victim Identification – overview to date, J. Forensic Odontostomatol. 23 (2005) 1–18. [3] T. Lay, H. Kanamori, C. Ammon, M. Nettles, S. Ward, R. Aster, C.R.S. Beck, S. Bilek, M. Brudzinski, R.M.R. Butler, H. DeShon, G. Ekstrom, K. Satake, S. Sipkin, The great

[4]

[5] [6] [7] [8]

[9]

[10]

[11] [12] [13]

[14] [15]

[16] [17] [18]

[19]

[20]

[21]

[22] [23] [24] [25]

[26]

[27]

97

Sumatra–Andaman earthquake of 26 December 2004, Science 308 (2005) 1127– 1133. P. Sribanditmongkol, P. Pongpanitanont, N. Porntrakulsree, M. Petju, S. Kunaratanapruk, P. Kitkailass, P. Ganjanarintr, N. Somboonsub, Forensic Aspect of Disaster casualty management Tsunami Victim Identification in Thailand, World Health Organisation, http://www.who.int/hac/events/tsunamiconf/presentations/ 2_16_forensic_pongruk_doc.pdf (accessed 18.06.14). NTERPOL DVI Guide, http://www.interpol.int/Media/Files/INTERPOL-Expertise/ DVI/DVI-Guide (accessed 18.06.14). INTERPOL DVI Forms, http://www.interpol.int/en/INTERPOL-expertise/Forensics/ DVI-Pages/Forms (accessed 18.06.14). J.P. Beauthier, E. Vlack, P. Lefevre, J. Winne, Mass Disaster Victim Identification: the tsunami experience, Open Forensic Sci. J. 2 (2009) 54–62. G. Lau, W.F. Tan, P.H. Tan, After the Indian Ocean tsunami: Singapore’s contribution to the international disaster victim identification effort in Thailand, Ann. Acad. Med. Singap. 35 (2005) 341–351. R. Lessig, C. Grundmann, F. Dahlmann, K. Rotzscher, J. Edelmann, P. Schneider, Tsunami 2004 – a review of one year of continuous forensic medical work for victim identification, EXCLI J. 5 (2006) 128–129. S. Ferreira, H. Kuser, R. Garrido, A. Trindade-Filho, K. Paula, M. Galva˜o, A. Moraes, Floods and mudslides in the State of Rio de Janeiro and a plane crash in the Brazilian Amazon rainforest: a study of two different experiences in disaster victim identification (DVI), Forensic Sci. Int.: Genet. Suppl. Ser. 3 (2011) e516– e517. H. Meyer, The Kaprun cable car fire disaster-aspects of forensic organisation following a mass fatality with 155 victims, Forensic Sci. Int. 138 (2003) 1–7. J. Chaikunrat, P. Pongpanitanon, M. Petju, Victim identification in the tsunami disaster in Thailand, J. Health Sci. 20 (2011) 897–902. G. Calacal, et al., Identification of exhumed remains of fire tragedy victims using conventional methods and autosomal/Y-chromosomal short tandem repeat DNA profiling, Am. J. Forensic Med. Pathol. 26 (3) (2005) 285–291. B. Olaisen, M. Stenersen, B. Meva˚g, Identification by DNA analysis of the victims of the August 1996 Spitsbergen civil aircraft disaster, Nat. Genet. 4 (1997) 402–405. M. Petju, A. Suteerayongprasert, R. Thongpud, K. Hassiri, Importance of dental records for victim identification following the Indian Ocean tsunami, Public Health 121 (2007) 251–257. M. Perrier, M. Bollmann, A. Girod, P. Mangin, Swiss DVI at the tsunami disaster: expect the unexpected, Forensic Sci. Int. 159S (2006) S30–S32. E. De Valck, Major incident response: collecting ante-mortem data, Forensic Sci. Int. 159S (2006) S15–S19. J. Kieser, W. Laing, P. Herbison, Lessons learned from large-scale comparative dental analysis following the South Asian Tsunami of 2004, J. Forensic Sci. 51 (2006) 106–112. M. Hennessey, World Trade Center DNA identifications: the administrative review process, in: 13th International Symposium on Human Identification, Promega, 2002, http://au.promega.com/products/pm/genetic-identity/ ishi-conference-proceedings/13th-ishi-oral-presentations/ (18.06.14). S. Donkervoort, S. Dolan, M. Beckwith, T. Northrup, A. Sozer, Enhancing accurate data collection in mass fatality kinship identifications: lessons learned from Hurricane Katrina, Forensic Sci. Int.: Genet. 2 (2008) 354–362. J. Dailey, Charting errors in mass disaster dental records; incidence, issues and implications, in: C. Bowers, G. Bell (Eds.), Manual of Forensic Odontology, 3rd ed., American Society of Forensic Odontology, Saratoga Springs, 1997, pp. 250–257. R. Byard, C. Winskog, Potential problems arising during international disaster victim identification (DVI) operations, Forensic Sci. Med. Pathol. 6 (2010) 1–2. M. Tsokos, R. Lessig, C. Grundmann, S. Benthaus, Experiences in tsunami victim identification, Int. J. Legal Med. 120 (2006) 185–187. L. Biesecker, et al., DNA identifications after the 9/11 World Trade Centre attack, Science 310 (2005) 1122–1223. B. Budowle, F. Bieber, A. Eisenberg, Forensic aspects of mass disasters: strategic considerations for DNA-based human identification, Leg. Med. 7 (2005) 230– 243. Z. Budimlija, M. Prinz, A. Zelson-Mundorff, J. Wiersema, E. Bartelink, G. MacKinnon, B. Nazzaruolo, S. Estacio, M. Hennessey, R. Shaler, World Trade Center human identification project: experiences with individual body identification cases, Croat. Med. J. 44 (3) (2003) 259–263. M. Prinz, A. Carracedo, W. Mayr, N. Morling, T. Parsons, A. Sajantila, R. Scheithauer, H. Schmitter, P. Schneider, DNA Commission of the International Society of Forensic Genetics (ISFG): recommendations regarding the role of forensic genetics for disaster victim identification (DVI), Forensic Sci. Int.: Genet. 1 (2007) 3–12.